In animal-pollinated flowers, the pollinators cannot detect the presence of nectar before entering flowers, and therefore flowers may cheat by not producing nectar. An earlier model suggested that a mixed strategy of producing nectarful and nectarless flowers would be evolutionarily stable. Here we compare nectarless flowers as a cheating strategy with three competing hypotheses namely "visit-more-flowers", "cross-pollination enhancement" and "better contact". We collected field data on 28 species of plants to test some of the differential predictions of the hypotheses. Nectarless flowers were detected in 24 out of 28 plant species. Correlations of percent nectarless flowers with floral and ecological variables support the cheater flower hypothesis. We further model the cost-benefits of cheating and show that an evolutionary stable ratio of nectarless to nectarful flowers can be reached. The equilibrium ratio is mainly decided by factors associated with pollinator density and pollinator learning.